History for BarryWarsaw

??changed:
-
Just some quick notes to capture some thoughts.

Background: I'm currently leading the development for 
"Mailman":http://www.list.org a mailing list manager written
Python.  Mailman has a bundled archiver called Pipermail, which
is really nice in that it mostly works, is written in Python,
and gives users a half-decent archiver out of the box.

Pipermail however has lots of problems: its code is very
difficult to maintain, it doesn't handle more complex email
types (i.e multiparts or attachments-- but MM2.1 will be
better!), it's not very efficient, and it doesn't support
searching, or more advanced forms of collaboration.

My short term goals would be to design a new archiver based
on Zope, centrally modeled around an email message type. 

A quick summary of features I want:

    - I want to throw all the messages into a big bucket
      and construct views on the archived messages on the 
      fly.  I don't
      like Pipermail's artificial separation of indexes into
      time-based buckets.  Too many threads span multiple
      months.

    - Everything's done through the catalog.  E.g. want to 
      find all the postings by me in the last 3 months?  Want
      to follow the entire *foobar* thread, even if it started
      6 months ago?  We should have canned catalog searches
      that mimic current behavior, but we can be much more
      flexible too.

    - Full-text searching.  Need I say more?

    - weblog.  I'm just starting to learn a bit more about
      weblogs and blogging, but I already know that I want
      followups and content to be addable from many differnt
      channels.  They should crosspost, much like Mailman's
      current news<-->mail gateway.

    - post via web.  Maybe this is weblog?  I want a button
      for anybody reading the archive via the web to be able
      to post a followup or reply to the current message.  I
      also want them to be able to start a new thread via
      the web.

    - forward via web.  See an article you like?  Or one your
      friend likes?  Fill in your email address (or not, if 
      you're already logged in) and hit the forward button.
      Note: this introduces an abuse vector, i.e. forward
      bomb people you don't like.  Possible mitigating factors
      include requiring a logon, message/rfc822 encapsulation
      daily limit on forwards to an email address.

      Note: you want to be able to forward a digest of 
      interesting messages in one swoop, but see above.

    - Dang it would be cool to provide an nntp front end
      to the archived messages.  Some people like reading
      discussions via email and/or digests, some like the web
      some like news.  We could (and should) cater to all.

    - attachment archiving.  All kinds of funky stuff comes
      in email.  we should strip out attachments of known
      types, compare them to various filters (e.g. 
      application/octet-streams get chucked), maybe pass them
      through virus checkers.  This needs a back door to the
      delivery process (what I call the "m-and-m", i.e. 
      moderate-and-munge) so that we could e.g. strip out
      large image attachments from messages exploded to the
      membership, replacing them with a url to the archived
      attachment.

    - url-based digests.  Mailman has two types of digests
      currently, RFC 1153 and RFC 934.  Both contain the
      full messages in their digests (with perhaps some 
      header trimming).  An alternative offering should be
      a summary digest which just contains urls to the
      articles in the archive.

Okay, so I admit I don't have a very lofty vision, but I
still think this stuff would rock!

Side note: I have other areas of interest for integration, or
leaching onto Zope technology.  I want to be able to integrate
membership databases with Zope (one membership in Zope, and you
manage your list memberships through it).  This would actually
be possible with Mailman 2.1.  I also want to explore the use
of page templates to replace Mailman's current extremely ad-hoc
web gui (some hardcoded scripts, some homegrown templating).